When you dig into what’s making automobile automation possible, you will find that a lot of computing horsepower is being packed into the design and it’s embedded into the vehicle itself. So why did the designers put the system into the car instead of in the cloud? Latency is critical, so the compute power must be tightly integrated into the control loop for the car. Like its pavement counterpart, you want to avoid a traffic jam on the information highway too. The last thing you’d want is for a round trip time to a cloud data center to prevent your self-driving car from safely delivering you to your destination.
This design model presents new challenges to us in the industry. There will be no single standard for edge IoT compute form factors. There will be common design points that most industry verticals will care about to some degree. Composing technology in a way that is adaptable to these various requirements will be a challenge that will separate the creative innovators from the fast followers. To help stimulate the innovators out there, let’s pick apart the edge IoT (Internet of Things) data center of the future with three basic design questions:
1. What workloads are we going to be running?
Machine learning and artificial intelligence are all the rage of late. These techniques aren’t new. What’s all the fuss about? In a word, it’s about data. We’re drowning in it. We’d prefer to be swimming. For many applications, we’ll need to have dynamic tuning of our sensor data as conditions change. This will require the embedded artificial intelligence driving the cars to go through training, re-training, and lots of it. This is the obvious answer. Beyond that, we’re going to need the infrastructure applications to connect our systems to users, the cloud, their peers, and potentially, the environment around them.
System designers need to grow in their thinking to incorporate these elements into their foundational designs for edge systems. Finally, they shouldn’t overlook the old-fashioned applications run in other places today. Putting data center-class computing power where it didn’t exist before will create incentive for software developers to build applications for it.
2. What are the primary constraints in the environment?
The answers here will vary, but nothing at the edge will behave like a data center does. Lack of strict environmental control and limited ventilation may be common. How many Gs do you experience when you hit a pothole at 60 kmph? Will we need airbags for our mobile IT? What is your high-availability requirement and model for redundancy? The idea of putting hyperconverged infrastructure into an edge appliance might raise questions, but it does make for a robust platform with a good deal of general compute capability.
3. How will we manage the lifecycle?
Data center IT is built based upon some principal assumptions of the lifetime of a generation of technology. These assumptions are wrong in many IoT environments.
As an example, you bought a Google TV in 2009. It’s a generation behind in display technology (HD instead of 4K), but it adequately delivers what you need for viewing most content nine years into its service. It will probably be in use for several more years. The brand stopped issuing updates for the embedded system years ago. The hardware simply no longer keeps up with new application development. Even if you wanted to, there is no way to upgrade the memory and CPU without replacing the entire television and its very expensive display. That makes it a technical orphan. It’s unfortunate. The idea was great. The execution failed to comprehend the difference in lifespan of the composed technology. We need to figure this out.
Beyond the hardware considerations, we also will need to design in a new model for management, maintenance and refreshing of our distributed data center technologies.
This is a new frontier in technology. The world is transforming because of it. We are setting the table. As we do so, we need to think beyond the technology packaging and delivery we do for traditional and cloud data centers today. Great innovations come from internalising the problems of the customer and applying technology in creative ways to solve them.
Are we thinking far enough from our comfort zone to make the leap to the future? Lenovo is driving edge computing to the center of our thinking about IoT and will be ready to meet you wherever your edge happens to be.